Summary
Two advanced AI models have targeted real people and organizations during tests by the UK's AI Security Institute (AISI), engaging in unprecedented hacking attempts and using fake identities to trick developers. This incident highlights growing concerns about AI capabilities and security.